Geochimica et Cosmochimica Acta, Vol. 65, No. 6, pp. 871– 884, 2001 Copyright © 2001 Elsevier Science Ltd Printed in the USA. All rights reserved 0016-7037/01 $20.00 ⫹ .00

Pergamon

PII S0016-7037(00)00576-7

The recovery and isotopic measurement of water from fluid inclusions in speleothems P. F. DENNIS,1,* P. J. ROWE,1 and T. C. ATKINSON1,2 1

School of Environmental Sciences, University of East Anglia, Norwich, NR4 7TJ, UK Department of Geological Sciences, University College London, London, WC1E 6BT, UK

2

(Received September 15, 1999; accepted in revised form October 16, 2000)

Abstract—The isotopic composition of speleothems is a useful palaeoclimatic indicator, but its value would be enhanced if information on the composition of the parent dripwaters could be recovered from fluid inclusions in the speleothem calcite. To develop a robust method for extracting and measuring oxygen and hydrogen isotopic composition of fluid inclusions we have used anhydrous Iceland Spar and microlitre glass capillaries of water as an analogue system. Crushing the capillary and calcite together in a high vacuum cell we have investigated the adsorbtive and isotopic behaviour of water when exposed to clean fracture surfaces. Significant water adsorption occurs at room temperature, accompanied by large negative isotopic shifts of both oxygen and hydrogen in the recovered free water at H2O/CaCO3 mass ratios ⬍10 mg g⫺1. Cryogenic pumping fails to achieve total desorption. The degree of depletion is inversely related to the water/calcite ratio, fractionation of hydrogen isotopes exceeding ⫺20‰, and oxygen isotopes ⫺10‰, at ratios typically observed in natural speleothems. Heating the crushed calcite at 150°C for 60 min. totally desorbs the water and allows retrieval of the correct isotopic composition. Application of these methods to a British Late Holocene speleothem yields ␦18O and ␦2H compositions for the inclusion water which are closely comparable with the modern cave dripwaters and local precipitation. The results show that isotopic compositions can be recovered from inclusion samples of ⬍1␮L (equivalent to approximately 1g of calcite) with precisions that are useful for palaeoclimatic research, ⫾0.4‰ for ␦18O and ⫾3‰ for ␦2H. Greater precision than this will require replicate analysis for each speleothem growth increment. Copyright © 2001 Elsevier Science Ltd and cave palaeotemperature. Both are important quantitative climatic parameters that can be modelled by, and therefore used to validate, GCM climate models. The importance of cave deposits as palaeoclimatic indicators has been recognised for nearly three decades (Hendy and Wilson, 1968) though, in the absence of drip water data, the climate-oxygen isotope link can be complex (Hendy, 1971; Gascoyne et al., 1981; Goede et al., 1986; Frumkin et al., 1999). Interpretation of temporal and areal trends in speleothem calcite composition are normally heavily reliant on other, independent, climatic evidence. A positive correlation exists between ␦18O in rainfall and condensation temperature (Dansgaard, 1964), but ␦18O in secondary calcite is negatively correlated with temperature of deposition (Kim and O’Neil, 1997). Quantitative interpretation requires calibration of the speleothem data against some sensitive indicator of local climate to establish which of these effects is dominant. Although many studies have adopted this approach (Harmon et al., 1979; Gascoyne et al., 1981; Goede et al., 1990; Dorale et al., 1992; 1998; Lauritzen, 1995; Bar-Matthews et al., 1997; Frumkin et al., 1999; McDermott et al., 1999), it is one that is not always possible. Complete characterisation of the palaeoclimate signal in speleothems requires data for the isotopic composition of the relict parent drip water that is preserved in fluid inclusions. Measurement of both the calcite and inclusion water leads to a direct deciphering of the climate signal, precluding the necessity for calibration against other proxy data. Pioneering investigations of inclusions in North American speleothems by Schwarcz et al. (1976), Harmon et al. (1978) and Harmon et al. (1979) were encouraging since they suc-

1. INTRODUCTION

In contrast to the high resolution marine record, the isotopic record of climate change in terrestrial deposits is less well documented and understood. Of the potential continental isotopic indicators of climate, speleothems (cave deposits) show great promise. They are widely distributed and speleothem deposition in low to mid-latitude regions occurred over long periods of the Quaternary. Through combining isotopic analysis with high precision uranium-series dating, reconstruction of long time series and time slices of regional climate variation at key stages in the recent geologic past becomes a realistic possibility. Speleothems are unique in containing a high resolution record of both palaeotemperature and isotopic composition of palaeogroundwater. They contain up to 0.1 wt.% of relict drip water in the form of fluid inclusions that were trapped in the host calcite at the time of growth (Kendall and Broughton, 1978). Modern cave seepage water is closely related to the mean annual meteoric precipitation at the cave site (Harmon et al., 1978; Yonge et al., 1985), while cave temperatures closely reflect the mean annual air temperature at the surface above the cave (Wigley and Brown, 1971; Gascoyne, 1992; Moore and Sullivan, 1978). Thus, provided that a speleothem grew in thermal isotopic equilibrium with its parent drip water and that there has been no subsequent re-equilibration, isotopic measurements of the inclusion and host calcite can be used to define the contemporaneous palaeoprecipitation isotope composition

*Author to whom correspondence should be addressed (p.dennis@ uea.ac.uk). 871

872

P. F. Dennis et al.

ceeded in detecting significant isotopic differences between inclusion waters of glacial and interglacial age. Similar studies by Rozanski and Dulinski (1987) succeeded in recovering the hydrogen isotopic composition of palaeodripwaters of interglacial ages from Poland. In these and other studies only the hydrogen isotope compositions of inclusion waters were measured. There has been a widely held view that fluid inclusions may have exchanged oxygen isotopes with the host calcite through geological time and thus ␦18O measurements may be of little value (Schwarcz et al., 1976; Rozanski and Dulinski, 1987). Whilst this proposition has not been systematically tested we have published preliminary data suggesting that a robust ␦18O signal can be recovered from some speleothem fluid inclusions (Dennis et al., 1998). In the absence of measurements, estimates of fluid inclusion oxygen isotope compositions have been made using present-day local meteoric water line relationships and the measured ␦2H composition (e.g., Goede et al., 1990). In addition to the lack of oxygen isotope data, later studies highlighted unresolved technical difficulties in the recovery and measurement of the hydrogen isotope composition of inclusion water with large deviations of analysed compositions from those expected. For example, Yonge (1982), Goede et al. (1986) and Goede et al. (1990) consistently found depletions of 20‰ to 30‰ in the hydrogen isotope composition of recovered modern inclusion waters relative to their associated cave dripwaters. Our own experiments, reported here, suggest that the measured depletion is an artefact of the analytical procedure, occurring as a result of isotopic fractionation during water adsorption onto crushed or calcined calcite surfaces. A deuterium and oxygen-18 depleted fraction is collected when recovery of the water is incomplete. These observations have tended to reinforce an underlying scepticism concerning the reliability of isotopic results derived from fluid inclusions, and since considerable accuracy (better than ⫾0.5‰ for ␦18O and ⫾5‰ for ␦2H) is required for useful climatic reconstruction, this must be demonstrated before inclusion isotopic measurements can gain wide acceptance. An additional constraint for inclusion studies is the desirability of high temporal resolution. Previous studies have generally used several grams of calcite for inclusion analyses, allowing only a very limited resolution. To resolve time scales of ⬍103 years requires development of methods for the recovery and measurement of inclusion waters with total volumes of less than 1 ␮L from subgram samples. In this paper we present the results of experiments with analog inclusion systems designed to help develop a robust technique for recovering the true isotopic composition of inclusion water from speleothems. The results highlight the importance of surface adsorption phenomena during liberation of inclusion water and the need for controlled thermal desorption to ensure 100% recovery without fractionation of either the hydrogen or oxygen isotopes. We have used the method to recover inclusion water from a British Late Holocene speleothem and present new results for the hydrogen and oxygen isotope composition that show high accuracy and precision with a temporal resolution of less than 200 yr.

2. DEVELOPMENT OF ANALYTICAL PROCEDURES FOR THE ISOTOPIC MEASUREMENT OF FLUID INCLUSIONS

2.1. Crushing Cell, Vacuum Preparation Line and Water Extraction Procedure To minimise the risk of oxygen isotope exchange between inclusions and the host carbonate, we have developed methods to crush rather than to thermally decrepitate or dissociate samples. Thermal decrepitation of inclusions in calcite at 400 to 500°C is accompanied by the release of large volumes of CO2 (Norman and Sawkins, 1987, Harmon et al., 1979; Schwarcz and Yonge, 1983; Yonge, 1982; Goede et al., 1986; Goede et al., 1990; Lecuyer and O’Neil, 1994). This hinders the efficient separation and isolation of the recovered inclusion water and may also lead to oxygen isotope exchange between the water and carbon dioxide. Previous inclusion studies using crushing methods have adopted off-line techniques whereby the sample container is evacuated and then removed from the vacuum line. The sample is then crushed, typically in a vice or in a ball mill (Harmon et al., 1979; Yang et al., 1995), after which the container is re-attached to the line and the released water cryogenically recovered. In contrast we have developed an on-line crushing cell and vacuum extraction line that allows dynamic cryogenic recovery of the water during the crushing stage. The cell design uses all metal gaskets and has no moving seals (Fig. 1). Initial experiments with a piston cylinder cell similar to the design of Andrawes and Gibson (1979) revealed problems due to inefficient crushing, incomplete water recovery, and an unacceptably high blank due to isotopically light water derived from the viton ‘O’-ring seals. The cell comprises a stainless steel tube welded to a NW35 conflat (CF) flange and bolted onto an NW35 CF base plate with an annealed copper gasket seal. The internal piston is a hardened steel mass that is suspended on a spring such that it rests just clear of the sample which sits on a hardened steel insert in the base plate. The piston is repeatedly lifted and dropped onto the sample by an electro-magnet with a frequency of ca. 2 Hz for as long as required. The cell is mounted on a hot plate and has a heating jacket that fits over the tube and piston to allow heating to temperatures of approximately 150°C. After loading with a sample the cell can rapidly be evacuated to a high vacuum and then crush the sample to silt size in a few minutes. Depending on the experiment the released inclusion water can be retained temporarily within the cell or collected immediately on-line in a cold trap at liquid nitrogen temperature. The vacuum line is illustrated schematically in Figure 2. It is constructed using 12 mm o.d. borosilicate glass tubing with Louwers-Hapert valves and Cajon ultra-torr connectors. To minimise water adsorption to either the glass or viton seals the assembly is wrapped with heating tape and the line kept at 80°C. Following crushing the released water is cryo-distilled into cold trap CT-A and held at liquid nitrogen temperature, for a period of up to 1 h. After the initial capture, the trap temperature is raised to ⫺120°C and any CO2 (typically about 0.2 ␮mols) and other non-condensable gases are pumped to waste. The manipulation of the trapped water then depends on whether hydrogen isotopes only are to be measured or those of both hydrogen and oxygen.

Isotopic composition of fluid inclusions

873

sealed 6 mm O.D. pyrex tubes. In both cases the water was reduced to hydrogen by reaction with the zinc at 500°C for 30 min (Coleman et al., 1982). For measurement of both ␦2H and ␦18O, the inclusion water is transferred from trap CT-A to an evacuated glass microequilibration tube at D and a carefully measured aliquot of CO2 added via the calibrated manometer (B), before flame sealing (Fig. 2). CO2-H2O micro-equilibration is carried out in a water bath at 25°C for a minimum period of 7 d (Kishima and Sakai, 1980; Rozanski et al., 1987). Following equilibration H2O and CO2 are cryogenically separated on trap CT-B (Fig. 2) which is held at ⫺80°C. The micro-equilibration vessel is loaded into a tube cracker attached to the vacuum line adjacent to the trap CT-B. The CO2 is transferred to either 6mm flame or valve sealed gas tubes for isotopic measurement, and the water then recovered into a Zn reduction tube as described above. Hydrogen isotopes were measured on a VG Micromass 602D isotope ratio mass spectrometer. The mass of water reacted is determined from the m/e ⫽ 2 signal for which the calibration curve is regularly checked to monitor drift. This is negligible between mass spectrometer overhauls. Carbon dioxide was measured on either a VG SIRA series II or a Europa Scientific Instrument Services (ESIS) SIRA mass spectrometer. Analytical precisions, based on replicate inclusion recovery and analysis, are ⫾2.7‰ and ⫾0.4‰ for ␦2H and ␦18O respectively. 2.3. Development of Artificial Inclusion-Calcite Systems as Speleothem Analogues

Fig. 1. The electro-magnetically operated crushing cell used for the elevated temperature water-calcite experiments and for extraction of the GB40 stalagmite fluid inclusions. The stainless steel tube is welded to an NW35CF flange, C, and bolted to a similar blanking flange. A hardened steel disk is located by the copper gasket. The hardened steel weight, A, is held clear of the sample by spring D. Two ports are provided to connect the cell to the vacuum line and also allow the use of a carrier gas to aid recovery of inclusion waters. In this study a carrier gas was not used.

2.2. Isotope Measurement Procedure For ␦2H measurement only, the water is cryogenically transferred from trap CT-A to a tube containing zinc granules at position D in Figure 2. In the early experiments we used acid-washed preheated BDH zinc shot (Coleman et al., 1982; Tanweer et al., 1988) in 12 mm O.D. pyrex tubes closed by Louwers-Hapert valves. Later, we used Indiana zinc in flame

To develop the recovery and measurement of ␮L water volumes from speleothems we made a series of experiments using a synthetic inclusion system. This consists of a single inclusion crushed with anhydrous Iceland spar with a ␦18O value of ⫺20.2‰VPDB. Artificial ‘inclusions’ were fabricated by taking up ␮L volumes of water in 5 ␮L capillary tubes which were then flame sealed. The internal laboratory standard NTW (Norwich Tap Water: ␦18O ⫽ ⫺6.84‰, ␦2H ⫽ ⫺44.9‰) was used in all these experiments. The mass of water in each capillary was determined gravimetrically and thus any loss of water or isotopic fractionation due to the experimental recovery procedures could be monitored. The capillaries more closely represent a single large inclusion, rather than the many micron-scale inclusions that are found in geological samples and are therefore not entirely analogous to the natural system we wish to investigate. However, we consider them to be sufficient to provide a reliable analog for likely isotopic behaviour. When inclusion-bearing rock samples are crushed, water is released progressively into an environment of freshly crushed grain surfaces, and our experiments simulate this quite closely. The cross-sectional size difference between the Iceland spar samples and the glass capillaries ensure that considerable calcite disintegration occurs before water release, thus providing a large active surface area (Fig. 1). Additionally the water released during crushing is allowed to adsorb onto the new surface area for a period of 10 min before cryogenic recovery. 3. ANALOGUE SYSTEM EXPERIMENTS

Results are available for recovered yields and isotopic compositions of water extracted in 3 types of experiment: (1) single

874

P. F. Dennis et al.

Fig. 2. Schematic diagram of the fluid inclusion recovery line. CT: cryotraps; A: crushing cell; B: baratron; C: CO2 reservoir; D: micro-CO2 cavity (oxygen measurement), or cold finger tube with Zn (hydrogen measurement). The line is constructed of borosilicate glass with Louwers-Hapert valves and Cajon ‘ultra-torr’ fittings. The assembly is wrapped in heating tape and maintained at 80°C.

glass capillaries crushed in the absence of calcite at room temperature; (2) single glass capillaries and calcite crushed at room temperature; and (3) single glass capillaries and calcite crushed at room temperature followed by heating to 150°C to thermally desorb any water bound to crushed calcite surfaces. 3.1. Results The results of experiments with single glass capillaries are presented in Table 1. The water recovery yields, lying in the range 98 to 104%, are within measurement error of 100%. The average oxygen isotope composition of the extracted water is ⫺6.84‰VSMOW ⫾ 0.42‰, and is identical to the calibrated composition of the laboratory standard. The average ␦2H composition of the water is ⫺46.8‰VSMOW ⫾ 3.1‰ and is within 2‰ of the accepted value. The data provide strong evidence that recoveries are essentially 100% with no observable blank contribution from the crushing cell and vacuum extraction line. This is further supported by the observation that the recovered water oxygen isotope compositions are identical with the accepted value of the laboratory standard NTW. The standard deviation of 0.42‰ is commensurate with direct measurements of ␮L aliquots of water using the H2O-CO2 micro-equilibration method and suggests that the extraction method does not contribute any additional variance to the data. There is, however, the possibility of a small bias in the hydrogen isotope measurements. The mean

hydrogen isotope composition of ⫺46.8‰ for the recovered water is depleted by 1.9‰ with respect to the calibrated value of ⫺44.9‰ for NTW. This is approximately 1.8 times the standard error of the mean and suggests that the bias may be real. The hydrogen isotope measurements were made using pyrex gas tubes sealed with a Louwers-Hapert valve employing viton ‘O’-ring seals. Results presented below for water release from capillary-Iceland spar systems and from a natural speleothem also show a small (2–3‰) depletion when compared to either the accepted NTW composition or to measurements on Table 1. Recovered yields and measured isotopic data for NTW sealed in glass capillaries and recovered from the crushing cell in the absence of calcite at room temperature. H2O Vol (␮L)

Yield (%)

␦18O ‰VSMOW

1.965 2.385 2.435 2.835 2.099 2.343 2.180 2.304 2.256 2.138 Mean values NTW composition

98 98 100 101 101 104 104 102 87 102 99.7 ⫾ 4.7

⫺7.59 ⫺7.04 ⫺6.43 ⫺7.10 ⫺6.97 ⫺7.16 ⫺6.78 ⫺6.48 ⫺6.63 ⫺6.18 ⫺6.84 ⫾ 0.42 ⫺6.84

␦2H ‰VSMOW ⫺52.3 ⫺47.1 ⫺45.7 ⫺44.3 ⫺48.0 ⫺48.9 ⫺45.8 ⫺49.4 ⫺41.2 ⫺44.8 ⫺46.8 ⫾ 3.1 ⫺44.9

Isotopic composition of fluid inclusions Table 2. Recovered yields and measured isotopic data for NTW sealed in glass capillaries and crushed with Iceland Spar. Recovery is at room temperature. Data for the BET specific surface area of crushed samples is also given. H2O/CaCO3 H2O Vol BET surface Yield ␦18O ␦2H (␮L) (mg⫺1 䡠 g⫺1) area (m2 䡠 g⫺1) (%) ‰VSMOW ‰VSMOW 9.30 6.76 6.54 6.53 4.96 3.80 3.77 2.50 2.28 2.12 1.89 1.21 1.15

4.55 3.40 5.19 3.35 2.21 2.44 3.15 1.29 2.53 2.72 2.72 1.75 1.51

0.65 0.55 — — 0.33 — 0.38 — 0.43 — — — —

100 100 96 100 86 77 84 70 87 93 86 51 46

⫺7.74 ⫺8.14 ⫺9.14 ⫺8.10 ⫺8.94 ⫺8.59 — ⫺10.11 — — ⫺11.04 ⫺16.79 ⫺17.04

⫺46.9 ⫺49.2 ⫺49.2 ⫺52.9 ⫺51.2 ⫺60.9 — ⫺64.9 — — ⫺60.4 ⫺66.9 ⫺68.4

875

replicate samples made using flame sealed tubes. Whilst developing the crushing cells we observed that at elevated temperatures viton releases small quantities of water that is isotopically depleted with respect to NTW. Thus we suspect that mixing of isotopically depleted water released from the viton seals with the sample during Zn-reduction is the primary cause of the small depletion observed. The results of room temperature experiments to characterise the yield and isotopic composition of recovered water from capillaries crushed with anhydrous Iceland spar are presented in Table 2. Also included in the data are measurements of the BET krypton adsorption specific surface areas of the crushed samples (de Kanel and Morse, 1979). The experiments use water to calcite ratios between 1.15 and 9.30 mg 䡠 g⫺1 (0.115– 0.93 wt.%) and overlap the range of water contents reported in previous speleothem inclusion studies (Harmon et al. 1979). The recovered water yields are plotted as a function of the water to calcite mass ratio in Figure 3. At ratios greater than approximately 6 mg 䡠 g⫺1 the yields are close to 100%. However towards lower ratios the yield reduces such that at 1mg 䡠 g⫺1 we recover just 50% of the initial water loaded into

Fig. 3. Recovery yield data for the synthetic inclusions plotted against water-calcite ratio for the room temperature (filled circles) and for 150°C desorption (open squares) experiments. The curved lines represent the yields expected for varying numbers of monolayers of water adsorbed, assuming a surface area of 0.50 m2 䡠 g⫺1, as measured by the BET method, and using an adsorption area of 10.6Å per water molecule.

876

P. F. Dennis et al.

Fig. 4. (a) Oxygen and (b) hydrogen isotopic compositions of recovered water from synthetic inclusions as functions of water-calcite ratio for desorption experiments at room temperature (filled circles) and 150°C (open squares).

the capillary. The general form of the yield versus water:calcite ratio plot is such that at water contents less than 1 mg 䡠 g⫺1 the expected yields reduce very rapidly towards zero. The oxygen and hydrogen isotopic composition of the recovered water are plotted as functions of the water:calcite ratio in Figure 4a,b respectively. All the experiments have produced recovered water with compositions that show a marked depletion with respect to NTW. The extent of depletion is a function of the water:calcite ratio. For ␦18O, depletions range from ⬇1‰ at ratios greater than 6 mg 䡠 g⫺1 to greater than 10‰ for ratios approaching 1 mg 䡠 g⫺1. A similar trend is observed for the depletion in ␦2H, ranging from 2‰ to nearly 25‰. At room temperature water is being retained in the crushing cell and is most likely adsorbing on to the fresh calcite fracture surfaces that are generated during crushing of the Iceland spar. These have measured specific surface areas of approximately 0.5 m2 䡠 g⫺1, Table 2. Adsorption is accompanied by a significant isotopic fractionation between the bound and recovered water. The magnitude of the fractionation effect is such that at all concentrations investigated up to 10 mg 䡠 g⫺1 the measured isotopic depletions in the

recovered water are unacceptably large for palaeoclimate and palaeoprecipitation studies. To improve the recovery of adsorbed water we investigated the effects of thermal desorption on the yield and isotopic composition. The results of these experiments are presented in Table 3. Capillaries of NTW were crushed with anhydrous Iceland spar at room temperature, the released water allowed to adsorb onto the calcite fracture surface, followed by heating of the cell to 150°C for a period of 60 min during which time the water was cryo-distilled over into the cold trap. Water:calcite ratios were selected towards the lower limit studied in the room temperature recovery experiments and cover the range 0.96 to 2.27 mg 䡠 g⫺1 (0.096 – 0.227 wt.%), Table 3. The recovery yields are plotted as open symbols in Figure 3. Yields are typically 100% with a minimum value of ⬎94%. These are in marked contrast to the room temperature extraction procedure which produced much lower yields for similar water:calcite ratios. The oxygen isotope composition of the recovered water, plotted as open symbols in Figure 4a, has a mean ␦18O value of ⫺6.49‰VSMOW ⫾ 0.26‰, Table 3. This represents an enrichment of 0.35‰ in 18O when compared to the accepted

Isotopic composition of fluid inclusions

877

Fig. 4. (Continued)

NTW composition of ⫺6.84‰VSMOW. There is also a weak trend towards heavier isotopic compositions with decreasing water:calcite ratios in the cell. The slight enrichment and trend might result from the fact that we made no attempt in these experiments to separate water from CO2 released during the crushing step. Later work showed that crush-

Table 3. Recovered yields and measured isotopic data for NTW sealed in glass capillaries and crushed with Iceland Spar. Recovery is at 150°C. H2O/CaCO3 (mg 䡠 g⫺1) 1.941 2.024 2.270 1.163 1.995 0.894 1.042 1.780 1.240

H2O Vol (␮L)

Yield (%)

␦18O ‰VSMOW

␦2H ‰VSMOW

1.90 ⬎98 ⫺6.55 ⫺47 2.12 ⬎94 ⫺6.43 ⫺49 2.41 100 ⫺6.80 ⫺47 1.72 100 ⫺6.17 ⫺47 1.84 100 n/m ⫺47 0.94 ⬎95 n/m ⫺47 1.57 ⬎96 n/m ⫺49 1.00 100 n/m ⫺47 1.00 100 n/m ⫺44 Mean values 98.1 ⫾ 2.5 ⫺6.49 ⫾ 0.26 ⫺47.1 ⫾ 1.5 NTW composition ⫺6.84 ⫺44.9

ing of the Iceland spar released small amounts (⬇0.1– 0.2 ␮mole) of CO2 with an enriched isotopic composition (⫹55‰VSMOW). Taking this small volume into account in the calculation of water isotopic compositions results in a correction of between ⫺0.1 and ⫺0.2‰ to the measured composition and draws the data to within 0.2‰ of the accepted value. In all later experiments with natural speleothems the recovered water was cryogenically separated from any CO2 released during the extraction procedure. The mean hydrogen isotope composition is ⫺47.1‰ ⫾ 1.4‰ with no trend in composition as a function of the water:calcite ratio, Fig. 4b. There is good agreement with the experiments on room temperature recovery of water from single glass capillaries in the absence of calcite, Table 1, as well as the accepted NTW composition of ⫺44.9‰. However, as in the experiments with single capillaries, there is evidence of a small systematic depletion of approximately 2‰ between the recovered water and the starting composition. 3.2. Discussion The initial experiments using glass capillaries in the absence of calcite show that the procedure of cryo-distillation of micro-

878

P. F. Dennis et al.

litre quantities of water from the crushing cell can be achieved with a 100% yield and no isotopic fractionation of the water. The observed 2‰ depletion in the measured hydrogen isotope composition results from the release of a small volume of isotopically light water from the viton seals used in the Zn reduction tubes and is not an artefact of the extraction procedure. However, when applying the same technique to capillaryIceland spar systems as analogues of natural inclusions, room temperature crushing and cryo-distillation results in poor yields and depleted isotope compositions for the recovered water. The data show systematic variations that can be semiquantitatively interpreted in terms of surface adsorption of the released water onto calcite fracture surfaces. Water vapour and other gas adsorption on freshly crushed mineral surfaces has been reported in fluid inclusion studies by several workers (Barker and Torkelson, 1975, Piperov and Penchev, 1973, Andrawes and Gibson, 1979). Superposed on the yields for room temperature recovery of water as a function of water:calcite ratio data (Fig. 3) are curves for the estimated yields assuming one, two, four and six layer adsorption. The curves were calculated using the mean BET measured specific surface area of 0.5 m2 䡠 g⫺1 for the crushed calcite, Table 2, and an adsorption area of 10.6 Å2 per water molecule (Gammage and Gregg, 1972). The general form of the yield versus water:calcite ratio data is very consistent with the adsorption model and suggests that between one and six layers of water remain adsorbed onto the calcite after cryo-distillation of the released water at ⫺196°C for a period of 40 min. This observation is consistent with the only previously published study of water adsorption onto calcite fracture surfaces by Gammage and Gregg (1972). Whilst measuring adsorption and desorption isotherms for water onto ball-milled calcite they report that after outgassing at 25°C for a period of 8 h a single monolayer of water remained bound to the calcite surface. Adopting the adsorption model the depleted isotopic composition of the recovered water, Table 2 and Figure 4, may be explained by isotopic fractionation between the bound and free water in the crushing cell. To account for the isotopic depletion fractionation factors for the partitioning of oxygen and hydrogen isotopes between the adsorbed and recovered water range between 1.01 to 1.03 and 1.05 to 1.15 respectively. Whilst there is a strong adsorption of water onto fresh calcite fracture surfaces at room temperature, quantitative recovery of the adsorbed water can be achieved using thermal desorption at 150°C combined with cryogenic recovery for a period of 60 min, Table 3 and Figure 3. This result is consistent with the adsorption study of Gammage and Gregg (1972) who report that water is rapidly removed from calcite surfaces during vacuum degassing at 150°C. Similarly quantitative recovery of fluid inclusion and adsorbed water at 150°C has been reported for silicate mineral-water systems (Andrawes and Gibson, 1979). During the desorption there is no evidence of fractionation associated with either the hydrogen or oxygen isotope compositions. The recovered compositions are in close agreement with the accepted value of the laboratory standard NTW, Table 3. At the desorption temperature of 150°C the isotopic composition of water in equilibrium with the Iceland Spar is approximately ⫺2‰VSMOW (O’Neil et al., 1969) and is sufficiently

different to the recovered water compositions to eliminate the possibility of oxygen isotope exchange between the crushed calcite surface and adsorbed water. Similarly, thermal decrepitation experiments with biogenic carbonate at 400°C suggest that there is no oxygen isotope exchange between the recovered inclusion water and the host calcite (Lecuyer and O’Neil, 1994). The results of these experiments provide a semiquantitative and robust description of the processes occurring during the low temperature crushing and recovery of water from an analog calcite-fluid inclusion system. We suggest that the variable depletion in the measured hydrogen isotope composition of speleothem inclusions reported in previous studies employing crushing methods results from the incomplete recovery of strongly adsorbed water (Yonge, 1982; Schwarz and Yonge, 1983; Goede et al., 1986; 1990). This is despite the use of different heating methods. More problematical are those studies where the inclusion water is released either through thermal decrepitation at moderate temperatures or thermal dissociation of the calcite (Yonge, 1982; Goede et al., 1986; Matthews et al., 2000). Again such studies have reported highly depleted hydrogen isotope compositions, by as much as 22‰ (Yonge, 1982) and 30‰ (Matthews et al., 2000). The origin of the depletions in these studies is not understood and requires further study. 4. MEASUREMENT OF FLUID INCLUSIONS IN A SPELEOTHEM

The stalagmite GB40 (Fig. 5) was collected from its growth position in GB Cave in the Mendip Hills near Bristol, England (lat. 51°18.2⬘N, long. 2°45.1⬘W; UK National Grid Ref. ST 47595623) in 1980. It has been U/Th dated and the part analysed here is known to have grown over the interval from c.5000 BP to almost the present day. This is a period in which variability in the palaeoclimate of Britain is small and difficult to detect from proxy records (Briffa and Atkinson 1997). Therefore, it may be expected that there was relatively little variation in the isotopic composition of palaeo-rainfall during the deposition of GB40, and that the palaeo-recharge waters trapped in fluid inclusions would reflect this. This is consistent with groundwater and paleogroundwater studies in UK aquifers (Bath et al., 1978; Hiscock et al., 1996). Our a priori expectation was that fluid inclusions in GB40 would show rather constant isotopic composition similar to that of the present day. In addition the fluid inclusion compositions could be compared with the unpublished results of two years’ isotopic study of dripwaters in this cave, which established that there is very little seasonal or interannual variation in their isotopic composition at the present day. The palaeoclimatic significance of GB40’s full isotopic record will be discussed in a future publication. Here we present only the isotopic measurements on fluid inclusions, and discuss their implications from a methodologic point of view. Figure 5a is a photograph of a slab cut radially across the stalagmite showing two major depositional units of which only the second is analysed here. The calcite of the upper unit shows strong, visible lamination which allows the growth axis and growth surfaces to be recognised for all stages of deposition. Figure 5c,d are optical photomicrographs showing thorn-

Isotopic composition of fluid inclusions

Fig. 5. (A) Photo of cut slab from stalagmite GB 40 analysed for fluid inclusions; (B) Drawn outline of slab (to same scale as (A) showing positions of fluid inclusion and dated samples (cross-hatched ornament). Samples used for ␦2H analyses are unshaded and singly hatched for ␦18O analyses. The layers are numbered sequentially from the top to bottom, 1 to 32 (see text and tables 4 and 5). All the samples for inclusion analysis were taken from the upper laminated unit (A) and straddle the growth axis; (C) and (D) Photomicrographs showing thorn shaped fluid inclusions within stalagmite GB40. The inclusion density is highly variable with the inclusions lying in bands (arrowed in (C)) parallel to the growth layering seen in (A). The scale bars are 20 ␮m long.

879

880

P. F. Dennis et al.

shaped fluid inclusions contained within columnar calcite crystals. At high magnification (Fig. 5d) smaller linear inclusions can be seen oriented along grain boundaries. The inclusion density is variable and Figure 5c indicates that the density variations define laminae, as first noted by Kendall and Broughton (1978). It is not known whether these correspond to the laminae visible in Figure 5a. For sampling purposes the slab of stalagmite was divided into 32 layers with boundaries parallel to the visible laminations and numbered from the top downwards. The location of samples is shown in Figure 5b. From a series of U/Th dates to be presented in the future paper the average time-span represented by each layer is 156 yr. Parallelopiped samples, typically 0.1 cm3 volume, were individually cut using a miniature diamond-wire band saw. Up to 9 replicates were cut from each layer within the radius of the shallow dome that formed the top of the stalagmite through most of its growth (Fig. 5a,b). It is expected that these along layer replicates will have identical fluid inclusion compositions, and the precision of our measurements of isotopic composition can be found by comparing them. For any given layer, a best estimate of inclusion composition can be found by averaging the values from several replicates, and a record of isotopic composition through time built up from successive layers. Ninety-four samples covering all 32 layers in the stalagmite were analysed for the inclusion ␦2H composition. These results are sufficient for a detailed analysis of errors and construction of an isotopic record. In contrast, only 15 samples were analysed for both ␦2H and ␦18O ratios. This smaller number is insufficient to establish a detailed ␦18O record, but is adequate to test the analysis methods. The two sets of samples are distinguished on Figure 5b and the results are discussed separately.

Table 4. ␦2H and H2O content data for GB40 (see Fig. 6b). F ⫽ flame sealed samples. All others are Viton sealed.

␦2H H2O H2O/CaCO3 ␦2H H2O H2O/CaCO3 Layer ‰VSMOW (␮L) (mg g⫺1) Layer ‰VSMOW (␮L) (mg g⫺1) 1

2

3

4

5 6 7

4.1. ␦2H Results from GB40 The full results are presented in Table 4, arranged by growth layer. From a methodologic point of view, we may approach them with two questions in mind, 1. What is the precision of the measurements? 2. How closely does the isotopic history of inclusion waters in GB40 conform to our a priori expectations? The analytical precision of individual measurements can be estimated from the variance of the differences between pairs of measurements. Provided that errors are random, the variance of differences is twice as large as that of the individual measurements. Results are available for hydrogen prepared in viton and flame sealed Zn reaction tubes. The viton sealed data provides nineteen wholly independent replicate pairs, and the flame sealed data gives seven. On this basis the standard deviation of viton sealed measurements is estimated as ⫾1.8‰ and that of flame sealed values is ⫾2.7‰. When viton sealed and flame sealed values from the same layers are compared, there are systematic differences between them. The mean of the viton sealed measurements is isotopically lighter than the corresponding flame sealed value in 22 layers out of 29. The Sign Test (Siegel and Castellan, 1988) indicates that such a disproportion has only a 0.4% chance of arising from random sampling error, so there appears to be a significant bias between the two methods. The mean difference

8 9 10 11 12

⫺48.0 ⫺52.5 ⫺55.0 ⫺53.0 ⫺49.5 ⫺53.0 ⫺53.0 ⫺51.0 ⫺44.5F ⫺49.5 ⫺50.0 ⫺50.0 ⫺49.5 ⫺42.5F ⫺39.0 ⫺47.5 ⫺52.0 ⫺57.5 ⫺45.0 ⫺47.0 ⫺49.0 ⫺43.5F ⫺43.0 ⫺47.5 ⫺47.0 ⫺38.5F ⫺38.5F ⫺45.0 ⫺50.0 ⫺44.0F ⫺51.5 ⫺45.0 ⫺51.0F ⫺49.0 ⫺45.0 ⫺41.0F ⫺46.0 ⫺48.0 ⫺48.0F ⫺50.5 ⫺47.0 ⫺48.0F ⫺46.5 ⫺45.0F ⫺46.5 ⫺40.0F ⫺49.0 ⫺45.0

0.35 0.25 0.30 0.45 0.40 0.30 0.30 0.25 0.25 0.27 0.25 0.35 0.27 0.45 0.30 0.40 0.35 0.30 n/m 0.27 0.26 0.32 0.42 0.35 0.40 0.40 0.30 0.35 0.37 0.28 0.40 0.43 0.53 0.45 0.50 0.30 0.50 0.65 0.45 0.40 0.50 0.50 0.55 0.30 0.70 0.50 0.60 0.65

0.88 0.79 0.90 0.80 0.79 0.83 0.68 1.05 0.72 0.81 0.82 0.85 0.74 0.84 1.00 0.88 1.08 0.95 n/m 0.62 0.78 1.09 1.06 1.04 0.97 1.02 0.97 0.90 1.21 0.92 1.27 1.14 1.70 1.28 1.25 1.25 1.37 1.23 1.78 1.47 1.38 1.42 1.56 1.79 1.53 1.55 1.89 1.54

⫺47.0 ⫺43.0F 14 ⫺44.0 ⫺44.0F 15 ⫺51.5 ⫺45.0 ⫺40.0F ⫺47.5F 16 ⫺46.5 ⫺45.0F 17 ⫺46.5 ⫺40.0 ⫺40.5F 18 ⫺46.0 ⫺42.0F 19 ⫺45.0 ⫺47.0F 20 ⫺47.0 ⫺47.0F 21 ⫺45.0 22 ⫺49.5 ⫺41.0F 23 ⫺49.5 ⫺39.0F ⫺47.0F 24 ⫺41.5 ⫺51.0 ⫺45.0 ⫺42.0F 25 ⫺45.0 ⫺46.5F 26 ⫺45.5 ⫺46.0F 27 ⫺43.5 ⫺42.0F 28 ⫺47.5 ⫺43.0F 29 ⫺51.5 ⫺38.0F ⫺47.5 30 ⫺48.0 ⫺41.0F ⫺43.0F 31 ⫺39.0F ⫺41.5F 31/32 ⫺50.0F 32 ⫺40.5F ⫺39.0F 13

0.65 0.60 0.55 0.55 0.40 0.53 0.70 0.55 0.65 0.66 0.55 0.70 0.30 0.60 0.43 0.53 0.60 0.53 0.40 0.80 0.60 0.40 0.75 0.65 0.70 0.50 0.45 0.40 0.59 0.45 0.50 0.40 0.74 0.60 0.69 0.40 0.60 0.65 0.75 0.48 0.45 0.56 0.75 0.50 0.70 0.50 0.77 0.35

1.57 1.50 1.98 1.65 1.74 1.71 2.10 1.91 1.57 1.89 1.70 1.54 1.32 2.05 1.37 1.92 1.58 1.82 1.65 1.92 1.71 1.87 2.16 1.91 1.88 1.55 1.72 1.71 2.33 1.79 1.78 1.22 2.05 2.75 2.45 1.75 2.24 2.12 1.72 2.03 1.94 n/a 1.88 1.96 1.67 2.07 1.51 1.40

is 2.72‰ with a standard deviation of 3.25‰. The standard deviation that would be expected from summation of the variances of the two methods is almost identical, 3.24‰, which lends support to the conclusion that experimental errors are in fact random as supposed. The speleothem data from the two methods may now be combined, by adding 2.72‰ to the viton measurements for each layer and calculating an overall mean. These combined estimates are listed in Table 5 and plotted in Figure 6. The errors quoted are standard errors of the mean for each layer, taking into account the individual variances of each measurement procedure and the additional variance contributed by the bias correction. Figure 6 thus constitutes a record of fluid inclusion composition in GB40 over the Late Holocene.

Isotopic composition of fluid inclusions

Whilst we are not considering the detailed palaeoclimatic interpretation of GB40, for the purpose of evaluating our methodology and procedures it is reasonable to define what we would expect to see in such a record and compare these expectations with the outcome. The Late Holocene was a time of limited climatic change, during which we might expect that the average isotopic compositions of precipitation and palaeorecharge would have remained fairly constant on century-tomillennial timescales. GB40 shows almost no significant deviation from an overall mean ␦2H of ⫺43.6‰, except possibly in the top two samples which are isotopically lighter and in the bottom two which are heavier. The record as a whole indicates relative constancy in isotopic composition during a period when palaeoclimate is believed also to have been rather constant. This is highly encouraging. Dripwater compositions were monitored for one or two years at 5 sites in GB Cave. Average compositions and standard errors are shown in Table 6 along with the averages for two sets of spot samples, for 14 sites on a single day in winter and 5 sites in summer. At ⫺43.6‰, the average value of ␦2H in stalagmite GB40 is close to the overall mean of the drip waters of ⫺45.1‰. The average ␦2H for every individual layer overlaps within experimental error with the range of drip waters. There is no obvious difference in ␦2H between the range of modern drip compositions and the measurements of palaeo-inclusion waters formed under similar climatic conditions during the Late Holocene. 4.2. ␦18O Results from GB40 Single samples were analysed for ␦18O composition from 15 layers, with results shown in Table 5. The data are matched with the previously measured average layer ␦2H compositions. The paired values are plotted in Figure 7 which compares them with the World Meteoric Line (WML) and the average compositions of drip waters and precipitation over GB Cave. The top two layers, 1 and 2, show markedly lighter ␦18O compositions than the main part of the stalagmite, in agreement with their somewhat light ␦2H ratios. These two layers plot within the field of seasonally-averaged modern drip water compositions, with the youngest layer in the centre of the field. Thus, the fluid inclusion compositions of the youngest layers accurately record both the ␦18O and ␦2H compositions of the drip waters in the cave. This study is the first in which the isotope ratios of both O and H have been recovered from speleothem fluid inclusions and demonstrated to reflect dripwater compositions. It is noteworthy that the drip waters themselves show variable composition. The values shown are seasonal averages based on regular sampling over one or two years, with standard errors of the mean. All of the drips show ␦2H ratios significantly heavier than the amount-weighted average for local precipitation (point ‘P’ on Fig. 7) while ␦18O ratios of drips are variable but with an average close to that of precipitation. The reasons for these variations are beyond the scope of the present paper but will be discussed elsewhere. The older layers of stalagmite GB40 show isotopic compositions which plot on or very close to the WML (Fig. 7). This is highly encouraging, as previous studies of cave drips by Yonge et al. (1985) in North America and Goede et al. (1986)

881

Table 5. Combined mean ␦2H and ␦18O data for GB40 (see the text and Fig. 6b). Combined mean and standard error (s.e.)

‘viton-sealed’ Growth layer

␦2H ‰VSMOW

‘Flame␦2H n sealed’ n ‰VSMOW

n

s.e.

1

⫺51.9

8 ⫺44.5 1

⫺48.7

9

1.53

2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32

⫺49.7 ⫺48.1 ⫺45.8 ⫺47.5 ⫺48.2 ⫺47.0 ⫺47.0 ⫺48.7 ⫺46.5 ⫺46.5 ⫺47.0 ⫺47.0 ⫺44.0 ⫺48.2 ⫺46.5 ⫺43.2 ⫺46.0 ⫺45.0 ⫺47.0 ⫺45.0 ⫺47.5 ⫺47.5 ⫺45.8 ⫺45.0 ⫺45.5 ⫺43.5 ⫺47.5 ⫺51.5 ⫺48.0 n.d. n.d.

4 7 3 2 2 2 2 2 1 1 2 1 1 2 1 2 1 1 1 1 1 1 3 1 1 1 1 1 1 0 0

⫺42.0 ⫺43.5 ⫺38.5 ⫺44.0 ⫺51.0 ⫺41.0 ⫺48.0 ⫺48.0 ⫺45.0 ⫺40.0 n.d. ⫺43.0 ⫺44.0 ⫺43.7 ⫺45.0 ⫺40.5 ⫺42.0 ⫺47.0 ⫺50.5 n.d. ⫺41.0 ⫺43.0 ⫺42.0 ⫺46.5 ⫺46.0 ⫺42.0 ⫺43.0 ⫺42.7 ⫺42.0 ⫺40.2 ⫺39.7

⫺46.0 ⫺45.1 ⫺41.2 44.5 ⫺47.3 ⫺43.2 ⫺45.5 ⫺46.7 ⫺44.4 ⫺41.9 44.3 ⫺43.6 ⫺42.6 ⫺44.6 ⫺44.4 ⫺40.5 ⫺42.6 ⫺44.6 ⫺47.4 ⫺42.3 ⫺42.9 ⫺43.9 ⫺42.8 ⫺44.4 ⫺44.4 ⫺41.4 ⫺43.9 ⫺44.7 ⫺43.1 ⫺40.2 ⫺39.7

5 8 5 3 3 3 3 3 2 2 2 2 2 4 2 3 2 2 2 1 2 3 4 2 2 2 2 3 3 2 2

2.05 1.62 2.05 2.65 2.65 2.65 2.65 2.65 3.25 3.25 2.63 3.25 3.25 2.30 3.25 2.65 3.25 3.25 3.25 3.72 3.25 2.65 2.30 3.25 3.25 3.25 3.25 2.65 2.65 1.91 1.91

1 1 2 1 1 1 1 1 1 1 0 1 1 2 1 1 1 1 1 0 1 2 1 1 1 1 1 2 2 2 2

␦18O ‰VSMOW ⫺8.22 ⫺7.56 ⫺8.44 ⫺6.91 ⫺6.83 ⫺6.67

⫺6.98 ⫺6.39 ⫺6.17 ⫺5.80 ⫺5.20 ⫺6.63 ⫺6.02 ⫺6.61

⫺6.78

in Tasmania have shown compositions close to the WML. It suggests that our inclusion analyses show no unexpected bias in either ␦2H or ␦18O ratios. The difference between waters from the older layers of the stalagmite and the top two layers suggests that there may have been a change in the isotopic composition of recharge in the last few centuries. This may have been related to climate but it is equally possible that the cause was disturbance to soil and rockhead by the trial pitting for lead and zinc that is evident on the ground above GB Cave. These aspects will be discussed in the future publication. 5. CONCLUSIONS

Simple synthetic inclusion experiments indicate that when fluid inclusions are ruptured at room temperature in the presence of freshly crushed calcite, there is preferential adsorption of isotopically heavier molecules onto active grain surfaces. At water concentrations of ⬍ 10 mg 䡠 g⫺1 of calcite, fractionation of the recovered water becomes significant. The degree of fractionation increases as the H2O/CaCO3 ratio decreases and it appears that complete desorption cannot be effected by cryogenic pumping alone. Inclusion water recovered in this way from natural calcite is likely to be isotopically depleted relative

882

P. F. Dennis et al.

Fig. 6. The average ␦2H composition of inclusions in each layer, plotted against layer number. The approximate age at the top and bottom of the record are indicated. The weighted mean average, and range of, modern drip water compositions in GB cave are indicated by the solid and dashed lines respectively. Layers for which both ␦18O and ␦2H data are available are indicated by an asterisk.

to its initial composition. The magnitude of the effect suggests that it may account for the negative deuterium shifts seen in some previous speleothem fluid inclusion work. The bound water can, however, be totally desorbed by heating to 150°C for 1 h, and its correct ␦18O and ␦2H composition recovered using CO2 micro-equilibration and zinc reaction techniques respectively, even at the low water concentrations normally found in speleothems. In addition, the very small amounts of inclusion water are susceptible to contamination by isotopically light water released from viton seals, and these should be eliminated from the recovery apparatus as far as possible. Assessment of these recovery and measurement procedures using a dated Late Holocene stalagmite from south west England, confirms their validity when applied to natural speleothem calcite. The comparison of isotopic measurements from GB40 with modern drip waters and the WML strongly suggests that fluid inclusions represent samples of the parent drip that deposited the speleothem. The methods of analysis presented here are capable of recovering the isotopic composition of inclusions in single samples with a precision of better than ⫾3‰ for ␦2H and ⫾0.4‰ for ␦18O. To detect and eliminate the gross errors which may occasionally affect analyses of such small quantities of water, replicate analyses of several samples from each layer are essential. Such replication has the advantage of increasing the precision of the estimated mean isotopic composition of each layer. Since the gross change in isotopic composition of midlatitude precipitation between glacials and interglacials may have been no more than

approximately 15 to 20‰ for ␦2H (Rozanski, 1985) and 2 to 3‰ for ␦18O (Bath et al., 1978; Stute and Talma, 1998, Stute et al., 1995), a high level of precision is essential for meaningful inferences about palaeoclimatic conditions to be drawn from speleothem inclusion data. Acknowledgments—We gratefully acknowledge the support of the University of East Anglia Research Promotion Fund in providing an initial grant to start this work and the continued support of the Natural

Table 6. Average ␦D and ␦18O compositions of dripwater samples from GB Cave.

Site

␦2H ‰VSMOW

␦18O ‰VSMOW

A. Seasonally monitored sites Blockhouse Drip ⫺45.8 ⫾ 1.4 (s.e.) ⫺9.1 ⫾ 0.30 (s.e.) Entrance Passage ⫺50.6 ⫾ 0.3 (s.e.) ⫺8.6 ⫾ 0.17 (s.e.) Entrance Drip ⫺46.0 ⫾ 0.4 (s.e.) ⫺7.9 ⫾ 0.15 (s.e.) Mud Passage Drip ⫺44.2 ⫾ 0.2 (s.e.) ⫺7.5 ⫾ 0.12 (s.e.) White Passage ⫺41.8 ⫾ 0.5 (s.e.) ⫺8.3 ⫾ 0.23 (s.e.) B. Sites sampled on a single day 14 sites (winter) ⫺45.1 ⫾ 2.4 (s.d.) — 5 sites (summer) ⫺43.1 ⫾ 1.0 (s.d.) — C. Weighted average ⫺45.2 ⫾ 0.21 (s.e.) — of all sites

Number of samples 9 9 30 24 12 14 5 103

s.e. ⫽ standard error of the mean; s.d. ⫽ standard deviation of sample values.

Isotopic composition of fluid inclusions

883

Fig. 7. ␦2H versus ␦18O plot of inclusion waters (solid circles), modern drip waters (D) and precipitation (P) for GB cave. Points (1) and (2) are the youngest layers of the speleothem (see text). Also shown is the World Meteoric Line (WML).

Environment Research Council under grants NERC GR3/9014 and NERC GR3/10473. Dan Moseley carried out the BET surface area measurements. The data on water isotopic compositions were obtained in collaboration with Peter Smart under grant NERC GR3/3177. We thank the perceptive reviews of Stephen Burns and two anonymous reviewers. REFERENCES Andrawes F. F and Gibson E. K. (1979) Release and analysis of gases from geological samples. Am. Mineral. 64, 453– 463. Bar-Matthews M., Ayalon A., and Kaufman A. (1997) Late Quaternary paleoclimate in the Eastern Mediterranean region from stable isotope analysis of speleothems at Soreq Cave, Israel. Quat. Res. 47, 155– 168. Barker C. and Torkelson B. E. (1975) Gas adsorption on crushed quartz and basalt. Geochim. Cosmochim. Acta 39, 212–218. Bath A. H., Edmunds W. M., and Andrews J. N. (1978) Palaeoclimatic trends deduced from the hydrochemistry of a Triassic sandstone

aquifer, United Kingdom. In Int. Symp. On Isotope Hydrology, IAEA, Vienna, pp. 545–568. Briffa K. and Atkinson T. C. (1997) Reconstructing Late-Glacial and Holocene climates. In Climates of the British Isles present, past and future. (eds. M. Hulme and E. Barrow), pp. 84 –111. Routledge, London. 454 pp. Coleman M. L., Shepherd T. J., Durham J. J., Rouse J. E., and Moore G. R. (1982) Reduction of water with zinc for hydrogen isotope analysis. Anal. Chem. 54, 993–995. Dansgaard W. (1964) Stable isotopes in precipitation. Tellus. 4, 436 – 468. Dennis P. F., Rowe P. J., and Atkinson T. C. (1998) Stable isotope composition of palaeoprecipitation and palaeogroundwaters from speleothem fluid inclusions. In Isotope Techniques in the Study of Environmental Change, pp. 663– 671, IAEA, Vienna. Dorale J. A., Gonza´lez L. A., Reagan M. K., Pickett D. A., Murrell M. T., and Baker R.G. (1992) A high resolution record of Holocene climate change in speleothem calcite from Cold Water Cave, Northeast Iowa. Science. 258, 1626 –1630.

884

P. F. Dennis et al.

Dorale J. A., Edwards R. L., Ito E., and Gonzalez L. A. (1998) Climate and vegetation history of the midcontinent from 75 to 25 ka: A speleothem record from Crevice Cave, Missouri, USA. Science. 282, 1871–1874. Frumkin A., Ford D. C., and Schwarcz H. P. (1999) Continental oxygen isotope record of the last 170,000 years in Jerusalem. Quat. Res. 51, 317–327. Gammage, R. B. and Gregg, S. J. (1972) The sorbtion of water vapour by ball-milled calcite. J. Coll. Int. Sci. 38, 118 –124. Gascoyne M. (1992) Palaeoclimate determination from cave calcite deposits. Quat. Sci. Rev. 11, 609 – 632. Gascoyne M, Ford D. C, and Scwarcz H. P. (1981) Late Pleistocene chronology and palaeoclimate of Vancouver Island determined from cave deposits. Can. J. Earth Sci. 18, 1643–1652. Goede A., Green D. C., and Harmon R. S. (1986) Late Pleistocene palaeotemperature record from a Tasmanian speleothem. Austr. J. Earth Sci. 33, 333–342. Goede A., Veeh H. H., and Aycliffe L. K. (1990) Late Quaternary palaeotemperature records for two Tasmanian speleothems. Austr. J. Earth Sci. 37, 267–278. Harmon R. S., Schwarcz H. P., and Ford D. C. (1978) Stable isotope geochemistry of speleothems and cave waters from the Flint RidgeMammoth Cave system, Kentucky: Implications for terrestrial climate change during the period 230,000 to 100,000 years B.P. J. Geol. 86, 373–384. Harmon R. S., Schwarcz H. P., and O’Neil J. R. (1979) D/H ratios in speleothem fluid inclusions: A guide to variations in the isotopic composition of meteoric precipitation? Earth Plan. Sci. Lett. 42, 254 –266. Hendy C. H. (1971) The isotopic geochemistry of speleothems—I. The calculation of the effect of different modes of formation on the isotopic composition of speleothems and their applicability as palaeoclimatic indicators. Geochim. Cosmochim. Acta 35, 807– 824. Hendy C. H. and Wilson A. T. (1968) Palaeoclimatic data from speleothems. Nature 219, 48 –51. Hiscock K. M., Dennis P. F., Saynor P. R., and Thomas M. O. (1996) Hydrochemical and stable isotope evidence for the extent and nature of the effective chalk aquifer of N. Norfolk, UK. J. Hydrol. 180, 79 –107. de Kanel J. and Morse, J. W. (1979) A simple technique for surface area determination. J. Phys. E: Sci. Instrum. 12, 272–273. Kendall A. C. and Broughton P. (1978) Origin of fabrics in speleothems composed of columnar calcite crystals. J. Sed. Petrol. 48, 519 –538. Kim S.-T. and O’Neil J. R. (1997) Equilibrium and nonequilibrium oxygen isotope effects in synthetic carbonates. Geochim. Cosmochim. Acta 61, 3461–3475. Kishima N. and Sakai H. (1980) Oxygen-18 and deuterium determination on a single water sample of a few milligrams. Anal. Chem. 52, 356 –358. Lauritzen S-E. (1995) High-resolution palaeotemperature proxy record for the last interglaciation based on Norwegian speleothems. Quat. Res. 43, (2), 133–146. Lecuyer C. and O’Neil, J. R. (1994) Stable isotope compositions in biogenic carbonates. Geochim. Cosmochim. Acta 58, 353–363. McDermott F., Frisia S., Huang Y., Longinelli A., Spiro B., Heaton T. H., Hawkesworth C. J., Borsato A., Keppens E., Fairchild I. J., van der Borg K., Verheyden S., and Selmo E. (1999) Holocene climate variability in Europe: Evidence from textural and extensionrate variations in three speleothems. Quat. Sci. Rev. 18, 1021–38.

Matthews A., Ayalon A., and Bar-Matthews M. (2000) D/H ratios of fluid inclusions of Soreq Cave (Israel) speleothems as a guide to the Eastern Mediterranean Meteoric Line relationships during glacial and interglacial times. Chem. Geol. 166, 183–191. Moore G. W. and Sullivan G. N. (1978) Speleology (2nd Ed.). Zephyrus Press, Teaneck, NJ. Norman D. I and Sawkins F. J. (1987) Analysis of volatiles in fluid inclusions by mass spectrometry. Chem. Geol. 61, 1–10. O’Neil J. R., Clayton R. N., and Mayeda T. K. (1969) Oxygen isotope fractionation in divalent metal carbonates. J. Chem. Phys. 51, 5547– 5558. Piperov N. B. and Penchev N. P. (1973) A study on gas inclusions in minerals. Analysis of gases from micro-inclusions in allanite. Geochim. Cosmochim. Acta. 37, 2075–2097. Rozanski K. (1985) Deuterium and oxygen-18 in European groundwaters–Links to atmospheric circulation in the past. Chem. Geol. (Isot. Geosci. Sect). 52, 349 –363. Rozanski K. and Dulinski M. (1987) Deuterium content of European palaeowaters as inferred from isotopic composition of fluid inclusions trapped in carbonate cave deposits. In Isotope Techniques in Water Resources Development. pp. 565–578, IAEA, Vienna. Rozanski K., Dulinski M., and Rzepka J. (1987) D and 18O analysis of milligram water samples. Appl. Radiat. Isot. 38, 713–716. Schwarcz H. P., Harmon R. S., Thompson P., and Ford D. C. (1976) Stable isotope studies of fluid inclusions in speleothems and their paleoclimatic significance. Geochim. Cosmochim. Acta. 40, 657– 665. Schwarcz H. P. and Yonge C. J. (1983) Isotopic composition of palaeowaters as inferred from speleothem and its fluid inclusions. In Palaeoclimates and palaeowaters: a collection of environmental isotope studies. pp. 115–133, IAEA, Vienna. Siegel, S. and Castellan, N. J. (1988) Nonparametric Statistics for the Behavioral Sciences. McGraw-Hill, New York, 399 pp. Stute, M. and Talma, S. (1998) Glacial temperatures and moisture transport regimes reconstructed from noble gases and O-18, Stampriet Aquifer, Namibia. In Isotope Techniques in Studying Past and Current Environmental Changes in the Hydrosphere and the Atmosphere. IAEA, Vienna, pp. 307–318. Stute M., Forster H., Frischkorn H., Serejo J. F., Clark J. F., Schlosser P., Broeker W. S., and Bonani G. (1995) Cooling of tropical Brazil (5°C) during the last glacial maximum. Science 269, 379 –383. Tanweer A., Hut G., and Burgman J. O. (1988) Optimal conditions for the reduction of water to hydrogen by zinc for mass spectrometric analysis of the deuterium content. Chem. Geol. (Isot. Geosci. Section). 73, 199 –203. Wigley T. M. L. and Brown M. C. (1971) Geophysical applications of heat and mass transfer in turbulent pipe flow. Boundary Layer Met. 1, 300 –320. Yang W., Spencer R. J., and Krouse H. R. (1995) Stable isotope and major element compositions of fluid inclusions in Devonian and Cambrian dolomite cements, western Canada. Geochim. Cosmochim. Acta 59, 3159 –3172. Yonge C. J. (1982) Stable isotope studies of water extracted from speleothems. PhD Thesis, McMaster University, Hamilton, Ontario, Canada. (unpubl.). Yonge C. J., Ford D. C., Gray J., and Schwarcz H. P. (1985) Stable isotope studies of cave seepage waters. Chem. Geol. (Isot. Geosci. Sect). 58, 97–105.